x

app.py

@@ -47,7 +47,7 @@ def calculate_similarities(image, *texts):
         if ttype=='hqclip':
             tokenized_texts = tokenizer(text_inputs).to(device)
         else:
-            tokenized_texts = torch.cat([tokenizer(text) for text in text_inputs]).to(device)
+            tokenized_texts = tokenizer(text_inputs,77).to(device)
         
         with torch.no_grad():
             image_features = model.encode_image(img_tensor)

debug.py

@@ -0,0 +1,27 @@
+import gradio as gr
+import open_clip
+import torch
+import numpy as np
+import pandas as pd
+from PIL import Image
+import matplotlib.pyplot as plt
+import io
+
+# --- 1. Model Initialization (run only once) ---
+print("Loading HQ-CLIP model...")
+model_hq, _, preprocess_hq = open_clip.create_model_and_transforms('hf-hub:zhixiangwei/hqclip-openai-large-ft-vlm1b')
+tokenizer_hq = open_clip.get_tokenizer('hf-hub:zhixiangwei/hqclip-openai-large-ft-vlm1b')
+print("HQ-CLIP model loaded.")
+
+print("Loading standard OpenAI CLIP model...")
+model_openai, _, preprocess_openai = open_clip.create_model_and_transforms('ViT-L-14-quickgelu','openai')
+tokenizer_openai = open_clip.get_tokenizer('ViT-L-14-quickgelu','openai')
+print("OpenAI CLIP model loaded.")
+
+device = "cuda" if torch.cuda.is_available() else "cpu"
+device='cpu'
+model_hq.to(device)
+model_openai.to(device)
+print(f"Models moved to {device}.")
+
+tokenizer_openai(['cat','dog'],77)